Algorithm and sort

[grepawksed]

Hi.
I have a vector with pointers to objects (vector<X*> ) that I would like to sort.
How can I do that with the sort that is in the STL algorithm package?

vector<X*> v;
sort(v.begin(),v.end(), .....)

but I'm not sure how I write the comparison function.
How do I do that?

[laserlight]

Something like this might work:

bool compareX(const X* a, const X* b) {
	return *a < *b;
}

// ...
vector<X*> v;
sort(v.begin(), v.end(), compareX);

[hk_mp5kpdw]

Assuming you've got the less-than operator (<) defined for your X objects, then it would be this...

class my_sort
{
public:
    inline bool operator()(const X* lhs,const X* rhs)
    {
        return *lhs < *rhs;
    }
};

...

sort(v.begin(),v.end(),my_sort());

That should sort them in ascending order.

[grepawksed]

Is it not possible to do without the extra class?

[laserlight]

Is it not possible to do without the extra class?

Yes, if you use my version, which uses a function pointer instead. A function object is better (e.g. more flexible) in some cases, especially if it inherits from std::unary_function or std::binary_function. I am not too sure of this myself, but compilers may decide to inline the comparison made via a function object, and may be less likely to do so for a function pointer.

[grepawksed]

Does your version also work if the class X is abstract?

It turns out that X is abstract and that the values of two X objects are calculated by the function

class X{
   virtual int value()=0;
};

that then should be implemented in a subclass.

so your compare function should change to
return a->value() < b->value()



Will this work? I'm getting the error
passing `const X' as `this' argument of `virtual
int X::value()' discards qualifiers

[laserlight]

If X::value() does not logically modify the state of X, then it should be a const member function.

class X{
   virtual int value() const = 0;
};

Alternatively, you can make bool operator< a friend function, and then implement the concept of "less than" as pertains to X.

[grepawksed]

Ah, great, it worked. Very big thank you!

[grepawksed]

Btw, can I skip the const? Even if it's better to have itperhaps.

[laserlight]

Btw, can I skip the const? Even if it's better to have itperhaps.

Yes, though I would be disturbed if the expression (a < b) modified the state of a or b.